E-Manufacturing: The Evolution Continues

It should be no surprise that a semiconductor manufacturing fab is loaded with computer power. At the fab level, a computer-integrated manufacturing (CIM) system is coupled with manufacturing execution software (MES) to track lots, manage recipes and monitor the status of tools.
 
What may be surprising is the lack of computer power when it comes to monitoring performance where it counts the most: at the most basic level of the process tool. Semiconductor manufacturing tools are complex entities and often need frequent maintenance or repair. Tool availability, or lack thereof, typically referred to as overall equipment effectiveness (OEE), can have a huge impact on fab efficiency. For the past five or six years, OEE has been identified in the International Technology Roadmap for Semiconductors (ITRS) as an area where major improvements are needed. To stay on the productivity curve defined by Moore's Law, experts say a 9-15% improvement in OEE per year is required. An estimate of the average for the overall equipment effectiveness for all equipment is only 40-50% (Fig. 1).

One of the effective tools available to achieve these goals is improved diagnostic capability. With advanced diagnostics, it's possible to detect potential failures well before they occur, order the necessary spare parts and, with the proper hook-up, even evaluate and fix problems remotely. With the right sensors in place, it should also be possible to characterize the tool so effectively that the need for test wafers is greatly reduced.

1. Equipment productivity (OEE) improvements are needed now more than ever. (Source: International SEMATECH)

These kinds of capabilities have been coined "e-diagnostics" with the idea that OEMs would be the ones best suited for monitoring their own tools over Internet-type connections. But e-diagnostics is really only one piece of a broader "e-manufacturing" puzzle that includes fault detection and classification (FDC), and advanced process control (APC), including run-to-run control (R2R) and real-time control.

International SEMATECH, working in conjunction with Japan's Selete and JEITA, has come up with a new term: equipment engineering. This refers to all operations for equipment availability improvement and performance maintenance inside and outside the factory. EE capabilities (EEC) may address any or all of the following areas:

• Line throughput maintenance and improvement
• Equipment health monitoring and troubleshooting
• Equipment performance improvement (especially newly introduced equipment)
• Collaboration with equipment suppliers (improvement, troubleshooting, redesign)
• Equipment parts, assembly versions, modification management
• Maintenance operation management and planning
• Process performance adjustment

"E-manufacturing is that process where fabs migrate to automated decision making based on data," said Brad Van Eck of International SEMATECH (Austin, Texas). "In order to reach that goal of automated data-driven decision making, which is obviously several years away, we are pushing back the impediments: the standards that are missing, the applications that are needed to analyze the data, and the infrastructure that is needed in fabs.

A prerequisite to an effective equipment engineering system (EES) is the ability to collect detailed data from the process tools. Figure 2 shows the distribution of the causes of equipment downtime, illustrating the type of data that needs to be collected. The top part of Fig. 2 shows average data for current factory equipment. Two-thirds of the downtime is associated with the equipment's "base functions." The bottom part of Fig. 2 is another analysis on equipment downtime averaged over newer technology equipment. Again, nearly two-thirds of downtime is associated with the "base functions."¹ The good news here is that, over the past several years, the semiconductor industry has gained experience with various field buses — also called sensor buses — used to collect data (see "It's Time to Board the Fieldbus Bus "). DeviceNet and Profibus are the most popular; Applied Materials, for example, is using DeviceNet on all of its new 300 mm tools.

2. These analyses show that work on equipment ‘base functions’ is critical to improve equipment productivity. (Source: International SEMATECH)

The interface of primary interest to the e-diagnostics community is called Interface A. One of the three standards that define Interface A has already been approved, and is now known as SEMI E125. This allows the equipment to define for the client application what kind of configuration it has, what software version is running, what messages it can provide, etc., so the application can know what can be tracked and diagnosed. The other two components of Interface A are up for approval at SEMICON West. "3507 has to do with security at the equipment level," said Harvey Wohlwend of International SEMATECH. "The next document, 3509, has to do with what we call data collection management, so you can control what data is collected, how often, based on certain kinds of events... those kinds of things."

The next priority is Interface C, which defines how data is moved from the server in the IC maker factory securely through the firewall back to the expert at the equipment supplier so that they can do remote diagnostics, while still protecting intellectual property. Interface B will address how various software applications running in the factory, such as APC software, share information.

Although the standards are still awaiting final approval, International SEMATECH has been working with various suppliers on prototyping tests. "We want to get prototypes of those standards early and shake out the bugs so that the suppliers get a tested version of the standard," Van Eck said. More details of the prototyping opportunities for suppliers can be found on the International SEMATECH web site, www.sematech.org. Many suppliers have also developed their own solutions in advance of the standards, and have actually had some form of e-diagnostics in place for several years.^2,3

Figure 3 is a diagram of a "dream" factory highlighting new EES components. Such a factory is probably years away, in that the move to e-manufacturing and e-diagnostics is likely to be much more evolutionary than revolutionary. However, there are definitely an increasing number of success stories. IBM and ILS Technology (Boca Raton, Fla.), for example, recently announced the successful installation of the first-of-its-kind, comprehensive e-diagnostics solution in IBM's 300 mm fab in Fishkill, N.Y. The solution, designed to enable remote tool monitoring, tool takeover and engineering collaboration under a single full-fab governance model, is now in full deployment after months of testing at IBM's fab.

3. Factories of the future are likely to incorporate the various equipment engineering capabilities shown. (Source: International SEMATECH)

Delays in e-manufacturing's evolution range from simple lack of funds during this severe downturn, to the lack of communication standards, and concerns over intellectual property security and increased pressure on the IT infrastructure. "2002 was a pretty tough year for e-diagnostics; there wasn't much money and people weren't really sold on it," said Russ Funk, iConnect product marketing manager for Brooks Automation (Chelmsford, Mass.). "But in 2003 we're starting to get more than a few visionaries investing in it. In the beginning, they get a lot of data and they're kind of inundated. It takes awhile for the OEMs and the fabs to work together to, as we say, put the 'diagnostics' in e-diagnostics." In one case, Funk said, a company has been able to give about 100 hours per year back to the fabs for each tool. "That's pretty good value."

Applied Materials has implemented web-enabled e-diagnostics on hundreds of tools, but do not try to link them to the fab's computer systems. "We have designed this so that it connects through a dedicated diagnostic interface on our tool and is completely non-intrusive with the existing customer's equipment control/fab control environment," said Rich Danielson of Applied Materials (Santa Clara, Calif.).

But the developers stress that it's less about connecting to the tool remotely and more about providing diagnostic capabilities. "Our charter is not to allow remote operation of the tool as much as it is just to understand how to get the tool into an on-line state from either a preventive maintenance state or unexpected down state," said Shawn Smith, general manager of yield management services at Applied Materials.

In terms of ROI, Danielson said that the system has been able to reduce parts and labor consumption. "We have seen a significant benefit in terms of faster and more effective diagnosis of tool/process problems. They are able to zero in and fix the right thing the first time and reduce labor consumption." Smith added that there are additional ROI benefits — it's just that they are less tangible. "There's the additional ROI that you will probably get in process performance, but it's hard to predict," he said. This would include something like chamber matching at a critical layer to improve yield.

"ROI has really been one of the Holy Grails of the whole e-diagnostic adoption," added John Field of Advanced Energy (Fort Collins, Colo.). "Everybody knows this is where we need to go for equipment productivity, but it's been slow coming off the ground because people don't want to pay for it. The problem with e-diagnostics is that it's very difficult to wrap a compelling, defensible ROI around that, even though we intuitively know that that's where we need to go."

Dave Hemker of Lam Research (Fremont, Calif.), who said that e-diagnostic capabilities have been built into the company's 2300 Series tools, agrees: "Now that we've provided the capabilities in the tool, how are we going to use them in a way that really pays off? Where is the return on the investment? That's the phase we're in. We're really trying now to develop the applications of this technology and then show the benefits to our customers."